Between January 1980 and July 1989, 63 total knee arthroplasties were performed on 48 patients who had a clinical history of diabetes mellitus. Six patients died prior to completing the 2-year follow-up period and two patients (3 knees) were lost, leaving 53 knees in 40 patients available for review. The patients were considered as a total group, and were also divided into two subgroups: (1) insulin-dependent diabetes mellitus (IDDM) (9 knees) and (2) noninsulin-dependent diabetes mellitus (NIDDM) (44 knees). There were 10 men and 30 women in the series with an average age of 67. Patients were followed from 24 to 126 months (average, 54 months). The Porous Coated Anatomic total knee system (Howmedica, Rutherford, NJ) was used without cement in 30 knees and with cement in 23 knees. Diabetic patients were matched to a nondiabetic control group using a computer-generated selection process to match the following parameters: (1) age at surgery, (2) sex, (3) arthritic diagnosis, (4) length of follow-up period, (5) mode of fixation, and (6) preoperative deformity. Statistical analysis was performed using regression analysis and chi-square tests. In the diabetic group, there was an average clinical score of 85, with 43 knees (82%) rated excellent or good. Ten knees (18%) were rated fair, poor, or required revision (4 revisions, 7%). This was statistically different from the control group, which had an average clinical score of 92, with 51 knees (96%) rated excellent or good, 1 fair rating, and 1 revision (P < .001).(ABSTRACT TRUNCATED AT 250 WORDS)